Model-based methods have recently shown promising for offline reinforcement learning (RL), aiming to learn good policies from historical data without interacting with the environment. Previous model-based offline RL methods learn fully connected nets as world-models that map the states and actions to the next-step states. However, it is sensible that a world-model should adhere to the underlying causal effect such that it will support learning an effective policy generalizing well in unseen states. In this paper, We first provide theoretical results that causal world-models can outperform plain world-models for offline RL by incorporating the causal structure into the generalization error bound. We then propose a practical algorithm, oFfline mOdel-based reinforcement learning with CaUsal Structure (FOCUS), to illustrate the feasibility of learning and leveraging causal structure in offline RL. Experimental results on two benchmarks show that FOCUS reconstructs the underlying causal structure accurately and robustly. Consequently, it performs better than the plain model-based offline RL algorithms and other causal model-based RL algorithms.

What if I would tell you that Language Models and Multi-Agent Reinforcement learning are now engaged and will get married soon? First and foremost, kudos to Andrés Fernández Rodríguez who sent me the inspiring paper "Multi-Agent Reinforcement Learning is a Sequence Modeling Problem". The idea of the paper is fantastic. In its essence, it is about mapping the problem of agent control to token translation. The authors use an encoder-decoder model like the original "Attention is all you need" paper.

The use of machine learning (ML) models in decision-making contexts, particularly those used in high-stakes decision-making, are fraught with issue and peril since a person - not a machine - must ultimately be held accountable for the consequences of the decisions made using such systems. Machine learning explainability (MLX) promises to provide decision-makers with prediction-specific rationale, assuring them that the model-elicited predictions are made for the right reasons and are thus reliable. Few works explicitly consider this key human-in-the-loop (HITL) component, however. In this work we propose HEX, a human-in-the-loop deep reinforcement learning approach to MLX. HEX incorporates 0-distrust projection to synthesize decider specific explanation-providing policies from any arbitrary classification model. HEX is also constructed to operate in limited or reduced training data scenarios, such as those employing federated learning. Our formulation explicitly considers the decision boundary of the ML model in question, rather than the underlying training data, which is a shortcoming of many model-agnostic MLX methods. Our proposed methods thus synthesize HITL MLX policies that explicitly capture the decision boundary of the model in question for use in limited data scenarios.

The combination of Reinforcement Learning (RL) with deep learning has led to a series of impressive feats, with many believing (deep) RL provides a path towards generally capable agents. However, the success of RL agents is often highly sensitive to design choices in the training process, which may require tedious and error-prone manual tuning. This makes it challenging to use RL for new problems and also limits its full potential. In many other areas of machine learning, AutoML has shown that it is possible to automate such design choices, and AutoML has also yielded promising initial results when applied to RL. However, Automated Reinforcement Learning (AutoRL) involves not only standard applications of AutoML but also includes additional challenges unique to RL, that naturally produce a different set of methods. As such, AutoRL has been emerging as an important area of research in RL, providing promise in a variety of applications from RNA design to playing games, such as Go. Given the diversity of methods and environments considered in RL, much of the research has been conducted in distinct subfields, ranging from meta-learning to evolution. In this survey, we seek to unify the field of AutoRL, provide a common taxonomy, discuss each area in detail and pose open problems of interest to researchers going forward.

We study lifelong reinforcement learning (RL) in a regret minimization setting of linear contextual Markov decision process (MDP), where the agent needs to learn a multi-task policy while solving a streaming sequence of tasks. We propose an algorithm, called UCB Lifelong Value Distillation (UCBlvd), that provably achieves sublinear regret for any sequence of tasks, which may be adaptively chosen based on the agent's past behaviors. Remarkably, our algorithm uses only sublinear number of planning calls, which means that the agent eventually learns a policy that is near optimal for multiple tasks (seen or unseen) without the need of deliberate planning. A key to this property is a new structural assumption that enables computation sharing across tasks during exploration. Specifically, for $K$ task episodes of horizon $H$, our algorithm has a regret bound $\tilde{\mathcal{O}}(\sqrt{(d^3+d^\prime d)H^4K})$ based on $\mathcal{O}(dH\log(K))$ number of planning calls, where $d$ and $d^\prime$ are the feature dimensions of the dynamics and rewards, respectively. This theoretical guarantee implies that our algorithm can enable a lifelong learning agent to accumulate experiences and learn to rapidly solve new tasks.

Deep reinforcement learning (DRL) is transitioning from a research field focused on game playing to a technology with real-world applications. Notable examples include DeepMind's work on controlling a nuclear reactor or on improving Youtube video compression, or Tesla attempting to use a method inspired by MuZero for autonomous vehicle behavior planning. But the exciting potential for real world applications of RL should also come with a healthy dose of caution – for example RL policies are well known to be vulnerable to exploitation, and methods for safe and robust policy development are an active area of research. At the same time as the emergence of powerful RL systems in the real world, the public and researchers are expressing an increased appetite for fair, aligned, and safe machine learning systems. The focus of these research efforts to date has been to account for shortcomings of datasets or supervised learning practices that can harm individuals.

Abstract: We discuss the problem of decentralized multi-agent reinforcement learning (MARL) in this work. In our setting, the global state, action, and reward are assumed to be fully observable, while the local policy is protected as privacy by each agent, and thus cannot be shared with others. There is a communication graph, among which the agents can exchange information with their neighbors. The agents make individual decisions and cooperate to reach a higher accumulated reward. Towards this end, we first propose a decentralized actor-critic (AC) setting.

This 3-course Specialization from Google Cloud and New York Institute of Finance (NYIF) is for finance professionals, including but not limited to hedge fund traders, analysts, day traders, those involved in investment management or portfolio management, and anyone interested in gaining greater knowledge of how to construct effective trading strategies using Machine Learning (ML) and Python. Alternatively, this program can be for Machine Learning professionals who seek to apply their craft to quantitative trading strategies. By the end of the Specialization, you'll understand how to use the capabilities of Google Cloud to develop and deploy serverless, scalable, deep learning, and reinforcement learning models to create trading strategies that can update and train themselves. As a challenge, you're invited to apply the concepts of Reinforcement Learning to use cases in Trading. This program is intended for those who have an understanding of the foundations of Machine Learning at an intermediate level.

Today we focus on building a Monte Carlo (MC) agent to learn a MDP. In a previous story, we implemented a model-based ADP learner which estimates a model of reward function r(s) and transition probabilities p(s′ s, a). This model-based approach may work efficiently in some cases. However, if the transition model is difficult to estimate, a model-free approach tends to be a better choice. Monte Carlo (MC), which is our topic today, is one example of such model-free approaches. The code in this story is part of our MAD from scratch project where MAD stands for machine learning, artificial intelligence, and data science. In model-based methods, our policy is derived from the utility values of states.

A longstanding goal of the field of AI is a strategy for compiling diverse experience into a highly capable, generalist agent. In the subfields of vision and language, this was largely achieved by scaling up transformer-based models and training them on large, diverse datasets. Motivated by this progress, we investigate whether the same strategy can be used to produce generalist reinforcement learning agents. Specifically, we show that a single transformer-based model - with a single set of weights - trained purely offline can play a suite of up to 46 Atari games simultaneously at close-to-human performance. When trained and evaluated appropriately, we find that the same trends observed in language and vision hold, including scaling of performance with model size and rapid adaptation to new games via fine-tuning. We compare several approaches in this multi-game setting, such as online and offline RL methods and behavioral cloning, and find that our Multi-Game Decision Transformer models offer the best scalability and performance. We release the pre-trained models and code to encourage further research in this direction. Additional information, videos and code can be seen at: sites.google.com/view/multi-game-transformers